Leaked: List of police, govt, uni orgs in Clearview AI's facial-recognition trials Clearview AI’s controversial facial-recognition system has been trialed, at least, by police, government agencies, and universities around the world, according to newly leaked files. Internal documents revealed by BuzzFeed News show that Clearview offered its technology to law enforcement agencies, governments, and academic …
The facial-recognition biz scraped billions of photos from public social media profiles, including Instagram and Facebook, and put them all into a massive database
Isn't that against the TOS and probably the law? If it's against the TOS then where is FB suing them to their last penny?
Terms of Service aren't necessarily legally enforceable - if it's genuinely publicly accessible (e.g. you don't have to log into an account to get the basic details) it would be rather hard to enforce: there has to be some sort of "tripwire" that requires you to agree to the ToS.
And Facebook would have to demonstrate an actual loss - at least in any sensible legal system.
Is it against the law? Well, that's presumably what the lawsuit will decide, at least in that specific jurisdiction. Other legal jurisdictions are available.
I presume that "can" means "will", as soon as the system is called upon to see whether a scanned image is already on the database it will automatically add it, if it isn't, and add to it if it is. In other words all of these guinea pigs is contributing to the database.
There's not much point in adding a random image to the database, since the whole point is to identify who the image is of. Without the identifying data an image is useless, and if you already had the data for the image you wouldn't need to look it up in the database :)
Context, refinement and joining up dots are important factors.
The timestamp of the capture, the location of the capture are useful (ask Bellingcat). For someone already in the database their appearance may have changed subtley - they will want to refine imagery for different angles of capture, different shades of light and for different grades of face stubble, or makeup, hair length, etc. If nothing else it arguably shows they were alive at that sighting. (Arguably, because someone, someday will try to smuggle Jeremy Bentham through Heathrow).
For someone not previously identified Cleartext would presumably try to guess the operator's motives in homing in on a specific target and to join up the dots. It could for example volunteer the information from other contemporaneous sightings that this unknown person took a strange route to the point they are at now. The fact that they are unknown does not preclude tracking, or being in some way a person of interest. Once identified, the information would naturally be entered into Cleartext as part of the documentation produced by the operator of some kind of security occurrence to facilitate further reporting. (Many year's ago I wrote a database system which did this in very rudimentary form compared to what is available today. It was used to provide statistics to justify the Security Department's budget for an NHS Trust).
"The study was criticized by bank and lender associations for not taking into account people's credit scores."
So the factor that helps determine if an applicant is actually likely to pay back the loan wasn't taken in to account?
The study is worth than worthless. Anyone with an iota of integrity wouldn't undertake and publicize the study in the first place without that information.
So, are you suggesting that black people with the same income and lifestyle and age and sex (+13 others parameters) as white people had *worse* credit scores?
Is it the melanin that makes black people bad at managing their credit or the curly hair?
At best your claim here, is that the credit scores are hiding the racism, rather than the mortgage lender.
Which, quite honestly, is probably correct. They almost certainly do have a race item in the scoring and will be double counting it. Since the researcher filters for one rather than the other, they'll be uncovering the double-counting based on pure racism.
e.g. credit company finds that [black] correlates to [poor housing] correlates to [mortgage arrears].
and that [poor housing] correlates to [mortgage arrears] and uses both parameters.
So technically [black] correlates to [mortgage arrears]. If however you control for the [poor housing] they're currently in, the credit score still contains the [black] double count.
Reading up, it seem's to be a feedback loop rather than double counting.
"But the financial activity that credit agencies use in scoring — mortgages, student loans, car loans and credit cards — is inherently biased against BIPOC borrowers, because it’s based on financial instruments that economic racism has kept out of the reach of racialized communities for decades."
https://medium.com/commonfuture/why-credit-scores-are-racist-da109fcfb300
Simply amplifying the racism each iteration.
I was in several photographs of groups years ago. Graduation ceremony photographs, wedding photographs. I did not consent for these photographs to be used for identification purposes (indeed there was no such thing as facial recognition software at the time). Actually there was no 'release' document signed by anyone, as I recall, so I don't even know who holds the copyright on these photos.
So what is the legal situation if someone decides to post 'their' graduation or wedding photograph on their social media account, identifying lots of other people (including me) and this is used by Clearview?
I would think that the legal situation is that you participated in an event where you knew your picture could be taken, so you implicitly consented to have your picture taken. Said picture is then the property of the person who took it, and that person may dispose of it as he or she intends.
I agree - but what if they e.g. tag it with your name linked to your face, and add some other personal identifying information they happen to remember? In that case an argument might be made they've stepped beyond any photo-copyright issues you may have given implied consent to.
Oh joy. If there are biases in AI, a group notably woke, in my experience, will have "fun" doing Improv with this. OTGH, it may be entertaining to see what monotonous drone the AI produces. It might make modern pollies sound competent by comparison. On second thoughts, nah.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
